DISCUSSION PAPER SERIES

DP14476

ENCOURAGING OTHERS: PUNISHMENTAND PERFORMANCE IN THE ROYAL

NAVY

Hans-Joachim Voth and Guo Xu

ECONOMIC HISTORY

LABOUR ECONOMICS

ISSN 0265-8003

ENCOURAGING OTHERS: PUNISHMENT ANDPERFORMANCE IN THE ROYAL NAVY

Hans-Joachim Voth and Guo Xu

Discussion Paper DP14476 Published 08 March 2020 Submitted 07 March 2020

Centre for Economic Policy Research 33 Great Sutton Street, London EC1V 0DX, UK

Tel: +44 (0)20 7183 8801 www.cepr.org

This Discussion Paper is issued under the auspices of the Centre’s research programmes:

Economic HistoryLabour Economics

Any opinions expressed here are those of the author(s) and not those of the Centre for EconomicPolicy Research. Research disseminated by CEPR may include views on policy, but the Centreitself takes no institutional policy positions.

The Centre for Economic Policy Research was established in 1983 as an educational charity, topromote independent analysis and public discussion of open economies and the relations amongthem. It is pluralist and non-partisan, bringing economic research to bear on the analysis ofmedium- and long-run policy questions.

These Discussion Papers often represent preliminary or incomplete work, circulated to encouragediscussion and comment. Citation and use of such a paper should take account of its provisionalcharacter.

Copyright: Hans-Joachim Voth and Guo Xu

ENCOURAGING OTHERS: PUNISHMENT ANDPERFORMANCE IN THE ROYAL NAVY

Abstract

Can severe penalties ”encourage the others”? Using the famous case of the British Admiral JohnByng, executed for his failure to recapture French-held Menorca in 1757, we examine the incentiveeffects of judicial punishments. Men related to Byng performed markedly better after hisunexpected death. We generalize this result using information from 963 court martials. Battleperformance of captains related to a courtmartialed and convicted officer improved sharplythereafter. The loss of influential connections was key for incentive effects – officers with otherimportant connections improved little after Byng’s execution or other severe sentences.

JEL Classification: D91, J20, N33

Keywords: Principal agent problems, punishment, Labor incentives

Hans-Joachim Voth - voth@econ.uzh.chUBS Center for Economics in Society and CEPR

Guo Xu - guo.xu@berkeley.eduHaas - UC Berkeley

Powered by TCPDF (www.tcpdf.org)

Encouraging Others: Punishment andPerformance in the Royal Navy∗

Joachim VothUniversity of Zurich

& CEPR

Guo XuUC Berkeley,

Haas School of Business

This version: March 7, 2020

Abstract

Can severe penalties ”encourage the others”? Using the famous case of the British Ad-miral John Byng, executed for his failure to recapture French-held Menorca in 1757,we examine the incentive effects of judicial punishments. Men related to Byng per-formed markedly better after his unexpected death. We generalize this result usinginformation from 963 court martials. Battle performance of captains related to a court-martialed and convicted officer improved sharply thereafter. The loss of influentialconnections was key for incentive effects – officers with other important connectionsimproved little after Byng’s execution or other severe sentences.

∗We thank seminar audiences at the universities of Melbourne and Zurich for comments, and Doug Allen,Sascha Becker, Peter Leeson, and Nicolas Rodger for helpful suggestions. Haoran Gao, Caitlyn Schwarzand Noah Sutter provided outstanding research assistance. Joachim Voth: voth@econ.uzh.ch; Guo Xu:guo.xu@berkeley.edu.

1

Introduction

”In this country, it is thought wise tokill an admiral from time to time toencourage the others.”

Voltaire, Candide

In 1756, a British fleet under the command of Admiral John Byng was sent to recapturethe Mediterranean island of Menorca from French forces, which had recently occupied it.After an indecisive naval battle, the British fleet withdrew to Gibraltar. In the court-martialthat followed, Byng was charged with neglecting his duty – he had failed to ”do his utmost”to recapture the island. In accordance with the Articles of War, he was convicted andsentenced to death. To the astonishment of almost all observers, King George II refusedto pardon Byng. He was shot on board HMS Monarch on March 14, 1757. Voltaire’s novelCandide describes the event in a fictionalized exchange: Candide asks ”And why kill theAdmiral?”; Martin replies ”...he did not kill a sufficient number of men himself... It wasproven that he was not near enough [the French Admiral]. In this country, it is thoughtwise to kill an admiral from time to time to encourage the others.” (Voltaire, 1759/2005) [ouremphasis]

The phrase ”pour encourager les autres” has passed into everyday parlance. It is typicallyused ironically, to argue that someone is being punished to spur greater efforts by others.While the phrase and concept are famous,1 there is little evidence that ”encouraging theothers” through exemplary punishments works – either in the 18th century Royal Navyor in general. Altruistic punishment may play an important role in sustaining cooperation(Fehr and Gachter, 2002). At the same time, punishment and other negative rewards cancrowd-out intrinsic motivation (Fehr and Rockenbach, 2003; Gneezy and Rustichini, 2004).When it comes to crime, harsher punishments should deter offenders (Becker, 1968).2

However, the evidence that poor prison conditions reduce crime rates is mixed, and thedisincentive effect of the death penalty is arguably weak (Chalfin and McCrary, 2017).3

In the military itself, there is also no clear support for the notion that ”encouraging theothers” results in better performance (Chen, 2016).

In this paper, we first examine the effect of Admiral Byng’s execution on the behavior ofBritish naval officers. Using newly-collected data on the combat performance of more than2,721 captains, we show that men with close ties to Admiral Byng fought much harder after

1Google lists 598,000 results for ”pour encourager les autres”, 6.8 million for ”encouraging the others”,and Google Scholar references more than 1,600 academic writings using the phrase.

2The idea that punishment serves a utilitarian purpose goes back to the philosophical and legal writingsof Bentham (1830) and Blackstone (1767/2017).

3See Katz et al. (2003); Mocan and Gittings (2006); Chen and Shapiro (2007); Donohue and Wolfers(2006). One important exception is evidence that greater strictness in sentencing influences criminal behav-ior by members of family and criminal networks (Bhuller et al., forthcoming).

2

his execution than before it. We then use newly-digitized data on 963 court martials of otherofficers to show that ”guilty” verdicts in general spurred greater action by those connectedto the officer in question. This holds both for family relations and professional ”followers”(i.e. men who had served with or under the court-martialed officer previously).

Two mechanisms could be at work. Either the punishment is more salient for those con-nected to the court-martialed officer, convincing men to take more risks and fight harder –perhaps because they feel they might be next. Alternatively, the loss of an influential pa-tron could mean that they have to rely more on performance (and less on connections) forcareer progression.4 To pin down mechanisms, we collect data on other patrons of treatedofficers. Captains who were connected to other influential officers did not change theirbehavior after a connected officer’s court-martial; but those that had no other obvious pa-tron suddenly began to distinguish themselves markedly more than before. Similarly, ifthe court-martialed officer with whom they were connected was more senior, subsequentperformance improved; but if he was junior to the officer in question, there was no change.These results suggest that ”encouraging the others” worked, but for reasons that are dis-tinctly different from the general deterrent effect implied by the phrase. Court-martialsof mentors pushed (genetically or professionally) connected officers with few other pa-trons to perform because they nullified the value of pre-existing connections, increasingthe incentives to win glory.5

1 Historical background

1.1 The Royal Navy and Naval Warfare during the Age of Sail

By 1815, the Royal Navy ruled the waves. From relatively modest beginnings in the 16thcentury, it had become by far the largest and most successful navy the world had ever seen,with close to a thousand ships sailing the seas, manned and maintained by over 125,000men at its peak. During the Napoleonic Wars, it was crucial for Britain’s eventual triumph.In naval battle after naval battle, it outfought its enemies, routinely inflicting losses on ascale of 7:1 or 10:1 on its enemies (Rodger, 2005; Allen, 2002; Robson, 2015). Between 1700and 1900, the Royal Navy’s victories were crucial in building the largest empire in world

4Some of the gain in performance may reflect a deliberate attempt to find new patrons.5This finding is similar to the results in Bhuller et al. (forthcoming), but documents the incapacitation

effect stemming from the loss of an influential connection.

3

history (Kennedy, 2010).

Many factors contributed to the Royal Navy’s success – but technological advantage wasnot one of them. British-built ships were often inferior to French, Dutch or Spanish vessels;guns were near-identical. Human capital, however, was often superior. Naval officers – incontrast to those in the army – often came from middle class backgrounds, and had topass exams before becoming captains. Noble blood – in contrast to the French Navy, orthe British Army – counted for little in promotion decisions (Rodger, 1986). Britain’s navycould also draw on large reserves of manpower in the merchant navy.

There is ample evidence that Royal Navy crews were more experienced, and better-disciplined,than those of other navies, showing superior seamanship and better gunnery (Allen, 2002;Robson, 2015; Rodger, 1986). Naval battles normally involved groups of battleships (“shipsof the line”) sailing close to each other. Such ships would carry from 50 to over 100 guns.In engagements at sea, often fought at distances of a few dozen yards, ships would usebroadsides (the simultaneous discharge of cannons on one side) until one of the vesselswas sunk or surrendered.

For an officer’s prospects, both performance and connections mattered. Boys with ”inter-est” (i.e. connections to high-ranking officers and admirals) would often start their careersas captain’s servants, aged 12 or so. They would then progress to become midshipmen, and- after an exam emphasizing navigation, seamanship, and gunnery - be promoted to lieu-tenant. The key step in a naval career was promotion to post-captain – the commanderof a large warship, of 20 guns or more (Pope, 2013; Rodger, 2005). Only about one thirdof lieutenants were eventually promoted. Once a post-captain, all assignments and pro-motions were principally based on seniority. Men with a poor reputation could, however,find themselves assigned to unattractive missions, sent to unhealthy islands, or on half-pay (i.e. as captains without a ship, receiving half their normal pay). Prior to promotionto post-captain, both employment and promotion prospects were highly uncertain. Perfor-mance mattered thereafter, too: Fighting prowess as a captain could accelerate promotionto commodore and admiral, as it did in the case of famous officers like Horatio Nelson.However, being connected to leading officers was also important for promotion (Voth andXu, 2020).

4

1.2 The Battle of Menorca

After the outbreak of the Seven Year’s War, French forces landed on Menorca. The Balearicisland was a key outpost of the Royal Navy. The British forces retreated into their fortress,St. Philip’s Castle. As soon as news of the landing reached London, the Admiralty dis-patched a British fleet under Admiral Byng. Byng was born in 1704, the son of an admiraland naval hero, and joined the navy aged 14. Aged 23, he was promoted to captain of aship with 20 guns. Thereafter, he commanded battleships for almost two decades, beforebecoming an admiral in 1745 (Robson, 2015; Baugh, 2008).

Upon receiving orders to relieve the Menorca garrison and retake the island, Admiral Byngfirst sailed to Gibraltar, and then to Menorca. There, the British and French fleets met onthe morning of May 20, 1756. From the outset, Byng had misgivings about the chances ofsuccess. The British had 13 ships of the line, the French 12. The British fleet manoeuvredinto an advantageous position, and the leading ships started to exchange fire with enemyvessels. However, many of the British ships failed to close with the French after one shipsuffered damage to its mast – following vessels stopped to avoid collision, and did not enterthe battle. Byng’s own flagship did not participate in the fighting at all (Dull, 2007).

The final outcome was that the French fleet withdrew – but Byng did not land the infantryintended to strengthen the Menorca garrison. At a council of war with his captains, hedecided to return to Gibraltar. Eventually, the British fortress garrison on Menorca surren-dered to the French. In Britain, there was public outrage at Byng’s failure to recapture theisland. Once in Gibraltar, he was arrested and sent to Britain. A court-martial in January1757 found him guilty of not “doing his utmost” to defeat the French. At the same time,the court recommended him to the mercy of the King. After extended public debate, theKing refused to commute the sentence, and Byng was executed aboard HMS Monarch onMarch 14, 1757.

1.3 Naval Discipline

In the Age of Sail, many engagements at sea were indecisive. In chance encounters betweenships, captains trying to avoid battle often pointed to their disadvantageous position. Inlarger engagements, lines of ships were slow to form up, and battles could be cut shortby the weather. Time and again, captains would decide not to engage because of concernabout being outgunned, short of shot, or in a poor position relative to the wind. Attempts

5

to evade battle were common enough for the leading historian of the Royal Navy to ob-serve:

”The [Royal] Navy during the Seven Years’ War had a real problem of cow-ardice. It was not so bad as it had been earlier in the century, but it remaineda source of intermittent anguish... and was several times the cause of seriousfailures in action... the three ranks in question were admirals, commanders andmasters.” (Rodger, 1986, p.244)

Discipline in the Royal Navy – like all other navies in the Age of Sail – was harsh. Cap-tains had extensive powers over their men. Floggings were a regular occurrence. Manyviolations according to the Articles of War carried the death penalty. Discipline for cap-tains themselves was also severe. In contrast to common practice in other navies, RoyalNavy regulations in general were carefully designed to force its captains to fight (Allen,2011). The Articles of War laid down a set of fighting instructions compelling commandingofficers to engage the enemy (see Appendix A.1 for an extract). British captains could becourt-martialed if they failed to attack enemy vessels as long as they were not clearly out oftheir own class. While frigates were not required to engage battleships, British ships wereexpected to attack and defeat enemy ships even if the latter had 50% greater firepower(Allen, 2002). Captains who failed to engage in a naval battle without good reason wereroutinely court-martialed, convicted, and never served again at sea.6

Court martials were common throughout the age of sail. Loss of ship – whether to theenemy or shipwreck – automatically resulted in a trial (Hannay, 1914). The court – com-posed of Royal Navy captains – would question witnesses, hear evidence, and examinedocuments before pronouncing its verdict. Most of the accused would be exonerated – areflection of a system stipulating automatic court martials for loss of every ship. Convictedcaptains could be sentenced to death. Most convictions resulted in milder sentences, butthe captain would routinely be ”broken” – he would lose his seniority, putting him at thebottom of the list of captains. In effect, he would be banned from ever commanding a navyship again.

At his trial, Admiral Byng was judged to have violated the Articles of War. Historians haveemphasized that the rules laid down in the Articles of War, and their stringent applicationto officers – including highly ranked ones like Byng – created

”a culture of aggressive determination which set British officers apart from their for-

6A good example is Captain John Williamson, of the Agincourt, who failed to engage during the battle ofCamperdown in 1797 (?).

6

eign contemporaries, and which in time gave them a steadily mounting psychologicalascendancy. ... British officers encountered opponents who expected to be attacked,and more than half expected to be beaten, so that [the latter] went into action withan invisible disadvantage which no amount of personal courage or numerical strengthcould entirely make up for (Rodger, 1986).”

1.4 An illustrative example

The case of Admiral Byng’s most prominent follower, his flag-captain Arthur Gardiner,illustrates how ”encouraging others” worked. Gardiner commanded Admiral Byng’s flag-ship Ramillies at the Battle of Menorca in 1756. He also gave evidence during the courtmartial, supporting Byng. Lord Anson, First Lord of the Admiralty, thought that bothByng and his officers had brought disgrace on the Royal Navy. After the guilty verdictand execution of Byng, Gardiner was in command of a smaller battleship, the Monmouth,with 64 guns. In 1758, his ship encountered a French vessel, the Foudroyant, with 80 guns(Rodger, 1986). To redeem his standing as a naval officer, and despite facing an enemywith almost twice the firepower of his own ship, Gardiner decided to attack.7 Gardinerhimself was shot and killed during the ensuing battle, but the Foudroyant eventually sur-rendered.

Captain Gardiner fought a much more powerful ship after Admiral Byng’s conviction. Thiscase demonstrates that, after Byng was made an example of, at least some of his followersfought harder. If connections and performance are complements, we should expect thisreaction – an exogenous negative shock to a captain’s position, influence, and prospects ofattractive postings and promotions will incentivize him to find other ways of improvinghis standing.

2 Data

Personnel and performance. To construct our officer-level panel, we draw on data fromThreedecks, a web resource featuring detailed information on vessels, crews, and navalactions. Threedecks constitutes the most comprehensive data source on the personneland ships of the Royal Navy and on fighting events during the Age of Sail. 8 Three-decks has assembled data on 25,229 ships, 33,959 seamen, and 1,022 actions and battles

7The Monmouth fired a broadside of 228 kilograms, whereas the Foudroyant had one of 440 kilograms.8The Royal Maritime Museum in Greenwich lists it as a standard reference.

7

among European sea powers (the major ones being British, French, Spanish, Dutch andPortuguese).9

Our core dataset is based on work by Voth and Xu (2020) and covers 5,848 British offi-cers (lieutenants, commanders, post-captains and admirals) and their assignments to 3,904Royal Navy ships during the years 1690-1849. Because we perform event studies, we limitthe sample to a balanced panel of officers around the court-martials of interest. Whenstudying the execution of Admiral Byng in subsection 3.1, for example, we transform thepanel into a high-frequency monthly panel, focusiing on the 206 officers active throughoutthe period 1756-58. For the analysis of all court-martials, the balanced panel covers 2,721officers in three year windows around 963 court-martial events.

Outcomes can be readily measured in this setting. Our dataset contains detailed informa-tion on the number of enemy ships captured or destroyed, the number of military actionsin which a ship participated, and whether the ship itself was captured, wrecked, or sunk.We follow Voth and Xu (2020) and combine the number of captures, ships destroyed, andactions joined into a composite measure of performance.10

Measuring connections. We measure connections using pre-determined family ties fol-lowing the methodology of Xu (2018). In brief, we match the officers with the Peeragedataset, a genealogical database that contains the family tree data of the British elite. Forany court-martialed officer, we compute the degrees of separation (by consanguinity) toall officers who are serving around the time of the trial. We define two officers to be con-nected if their degrees of separation is less than 16. This is the same cut-off used in Xu(2018) and Voth and Xu (2020). Results are robust to using alternative cut-offs.11 Finally,we also leverage the rich personnel data to construct a measure of on-the-job ties. We de-fine two officers to share a tie if they have previously served on the same ship in the sameyear.12

Table 1 provides descriptive statistics for the 206 officers for which we have data around theexecution of Admiral Byng. We compare the differences between Byng’s kins to non-kinspre-execution (up to 1756) and post-execution (1757-1758). By our measure, 19% of offi-

9These numbers are correct as of September 2018 – the date when we downloaded the relevant data.10To validate this measure, Voth and Xu (2020) show that higher performing officers are promoted more

quickly to post-captain.11Xu (2018) contains a detailed discussion of the construction and interpretation of this measure. We later

test the robustness of our findings with respect to this measure.12Ager et al. (2016) use the same definition to establish ties between WW II fighter pilots.

8

cers are connected to Admiral Byng. Given his aristocratic background, this is a reasonableshare.13 On average, Byng’s kins and non-kins have served a comparable number of years.Prior to his execution, they also do not differ significantly in their capture rate, the numberof enemy ships sunk and actions participated in, or the combined measure thereof (bat-tle performance). Post-execution, however, Byng-connected officers exhibit significantlyhigher capture rates and outperform unconnected kins. Connected captains commandmore frequently and are assigned slightly more powerful vessels before Byng’s trial, butthis is no longer true post-execution. Due to Byng’s seniority and influence, his kins, how-ever, hold higher positions in the navy: those connected to Byng are 18.3% points morelikely to have become post-captains (i.e. be in independent command of warships).

Court-martials. We systematically collected data on court-martialed officers. For eachofficer in our dataset, we searched the Oxford Dictionary of National Biography as wellas John Marshall’s Royal Naval Biography – the authoritative historical compilation of navalcareers – for mentions of court-martials. For officers without entries, we conducted anextensive online search. Overall, we identify 186 court-martials (including the famouscase of Admiral Byng). In 53 of the cases, the court-martialed officer was convicted. Tothese we add the large number of automatic court-martials due to loss of a ship.14

The resulting total number of court-martials is 963. Since we are interested in studying thespill-over effects of the trials on the relatives of court-martialed officers, the main analysisrestricts the total number of court-martials to the subset for which we can compute thekinship networks. The resulting dataset contains 168 trials, covering the court-martials ofwell-connected and more prominent officers. We leverage the full sample when general-izing the results using on-the-job ties as an alternative measure of connections.

13Byng’s father was George Byng, ennobled as the 1st Viscount of Torrington. He served as Admiral of theFleet and Treasurer of the Navy, and was twice elected as the Member of Parliament for Plymouth.

14We assume that, in the absence of information to the contrary, these resulted in acquittals. This assump-tion does not change our results, and we also report in Table A6 the results without the acquittals followingthe loss of a ship.

9

3 Results

3.1 Encourager les autres - Admiral Byng

To examine if others were encouraged by punishments, we compare the performance ofconnected and unconnected officers, before and after a guilty-verdict and/or punishmentby a court-martial in the Royal Navy. Figure 1 illustrates the effect in the case of AdmiralByng.15 Prior to March 1757, naval officers related to Admiral Byng did not outperformtheir peers. Immediately after his execution, however, the performance of connected offi-cers spiked, increasing by a factor of 10 for three months. While the differential declinesthereafter, it remains sizeable for the rest of the year. As we know from the case of CaptainGardiner, tainted by his association with Byng, some officers even two years later tried toredeem themselves.

We perform a simple balanced difference-in-difference analysis for the period 1756-58 toquantify the magnitude and establish the statistical significance of our findings. Specifi-cally, we estimate for outcome yit of officer i in year-month t:

yit = β × Byng kini × Postt + θi + τt + γ′xit + εit (1)

where yit is the combined number of captures, ships sunk and actions participated in. Thedummy Byng kini is 1 if the officer is connected to Byng.16 Postt is a dummy that is 1 in themonth following Admiral Byng’s court-martial and execution. θi are officer fixed effectsand τt are year-month fixed effects. xit is a vector of individual controls which we willdiscuss subsequently. We cluster the standard errors at the officer-level.

Table 2 shows the regression results. Byng-connected officers do not outperform beforeByng’s execution – nor is there a general change in performance after March 1757 (column[1]). For connected officers, however, the average performance per month is suddenly 64%higher than average. The results are similar if we control for year and officer fixed effects(columns [2] and [3]). As Table 1 shows, officers connected to Byng are more elite andhigher ranked. While the individual fixed effects absorb such time-invariant differences,

15The coefficients are estimated using a flexible version of the specification in Table 2, column [3] wherewe allow the effect of the connectedness to Byng to vary for each month.

16The results are also robust to transformations of the dependent variable such as a dummy for any capture,ships sunk or action, and the inverse hyperbolic sine transformation to address the large number of zeros.We verified the robustness of our results using negative binomial count models (Table A1).

10

these level differences could nonetheless interact with the execution of Admiral Byng. Incolumn [4], we interact all baseline covariates from Table 1 with the Postt dummy. Theresults remain virtually unchanged. Finally, we also consider the possibility that the mea-sure of connectedness to Byng is – at least in part – capturing the general connectednessto the elite. To that end, we limit the sample to only those officers who are listed in theelite Peerage dataset (column [5]). Assuringly, we find the same results within the sampleof the British elite. In other words, there is no evidence that those born with high statusin general felt that their standing and status had now suddenly become more insecure,leading to greater efforts. Instead, the change in performance is specific to men related toAdmiral Byng.

Robustness. The results are robust to different cut-offs for defining connectedness to Ad-miral Byng. In Figure A2, we report the estimates of Table 2 for cut-offs ranging from 12 to20 degrees of separation. As expected, the effect declines with the degrees of separation.The results, however, remain statistically and economically significant throughout. Wealso provide robustness checks to ensure that the results are not driven by confounders.In theory, the greater performance of Byng-connected officers could be driven by more fa-vorable assignments – either to easier tasks or to more powerful ships. This, however, isunlikely – relatives of an Admiral executed for cowardice would hardly be favored in theirsubsequent assignments. As Table A2, Panel A, column [1] indeed shows, kins of AdmiralByng are statistically not more likely to command a ship post-execution than non-kins; fur-thermore, conditional on being assigned a ship, kins of Admiral Byng command equallypowerful vessels, as measured by their gun count (column [2]). Allowing the performancedifference between Byng’s kins vs. non-kins to vary by being given command thus doesnot affect the estimates (column [3]).

Another concern is that the observed effect could be driven by the Seven Year’s War (1756-1763), which falls into the time period of our study. Of course, for Royal Navy captainsto fight successfully, the country had to be at war. But it could be that the inital bout ofnaval fighting biases results in our favor. While the fact that the effect only appears inthe immediate aftermath of the court-martial helps alleviate such concerns, we go a stepfurther and also control for monthly variation in the overall conflict intensity during thewar. As Figure A1 shows, conflict events – as measured by the total number of enemyship captures, ships sunk and actions participated – spike during wars but there existssubstantial within-war variation. In Table A3, column [1] we thus allow the performancegap to vary by the total number of conflict events in a month. Reassuringly, the results do

11

not change substantially, suggesting that the effect is indeed driven by connections withAdmiral Byng.

3.2 Court-martials and the encouragement of others

Is “encouraging others” only effective in famous cases like that of Admiral Byng? Byngwas by far the most high-ranking officer to be convicted (and executed) in British navalhistory – and his convinction came as a surprise. It is an open question whether similareffects could follow from the conviction of less-elevated officers. To go beyond a single casestudy, we draw on all Royal Navy court-martials during the period 1690-1849. This alsoallows us to go beyond kinship connections, and to look at broader measures of social ties –e.g. whether two officers have previously served on the same ship. Overall, we identify 963court-martial trials. While this allows us to generalize the findings, a drawback of usinglesser known cases is that details about the trials are less abundant. For example, we donot have the exact dates of all court-martial trials, allowing us to only conduct the eventstudy at an annual level.17

Empirically, we proceed as before, comparing relatives to non-relatives before and after thecourt-martial trial. We pool all court-martial cases by ”stacking” each event study, whichwe confine to a balanced 3 year window around each trial. For court-martial case c, officeri and the year around the court-martial event t = {−3,−2, ...3}we estimate:

yict = β × Kinic × Postct + γ × Kinic × Postct ×Guiltyc + θic + τct + κK(i,t) + εict (2)

where Kinic is now a dummy if the officer i is connected to the court-martialed officer c.Postct is a dummy that is 1 in the year of the court-martial trial and thereafter (t ≥ 0).Guiltyc is a dummy that is 1 if the court-martial c resulted in a guilty verdict, and 0 other-wise. This distinction is important as the average effect of a court-martial trial will masksubstantial heterogeneity. While Admiral Byng was found guilty and executed, the major-ity of court-martial cases ended in acquittals, providing an additional source of variation.18

17As Table A4 shows, however, the effects of Admiral Byng’s execution also hold on the yearly-level. Thisalleviates concerns that the level of analysis might affect the results.

18Convictions are an endogenous outcome, but we do not find much evidence that – conditional on a court-martial – those connected to guilty officers look substantially different from acquitted officers (see Table A5.),other than that they tend to command smaller ships and are less likely to have lost a ship (most officers wholost a ship were acquitted).

12

Since officers can be exposed to multiple court-martial trials, we constrain the comparisonto each officer × court-martial trial using the officer × court-martial FEs θic. τct are court-martial specific fixed effects for each year around the court-martial trial. κK(i,t) are fixedeffects denoting the years of tenure k = K(i, t) for officer i in year t. We cluster the standarderrors at the officer × court-martial level.

The regression results are reported in Table 3. Kins of court-martialed officers outperformpost-trial when the accused officer was found guilty (column [1]). In contrast, kins ofcourt-martialed officers underperform when the accused officer was acquitted.19

Next, we go beyond the kinship connections by looking at on-the-job ties that were es-tablished through officers serving on the same ship. In the Royal Navy, the relationshipbetween the captain and the first lieutenant was a key pathway for patronage.20 The use ofon-the-job ties also has the added advantage that we can extend our sample substantially,going beyond those for whom peerage data on family ties is available. As column [2]shows, our main results are confirmed using on-the-job ties as an alternative measure ofsocial connections. The size of these effects are large, but smaller than in the case of Admi-ral Byng. There, the guilty verdict led to an average performance increase by 118%, relativeto the mean of the dependent variable.21 Here, the average gain for all court-martials forrelatives was 20%, and 50% for professional ”followers”. Finally, we combine both mea-sures – connectedness by kin and previous service – to construct a composite measure thattakes the value of unity if the officer was either a relative or had previously served with theofficer. As column [3] shows, the results remain comparable. As before in subsection 3.1,we provide robustness checks to ensure that the results are not driven by a more favorableassignment for kins of guilty officers post-execution (Panel B, Table A2, columns [1]-[3],or confounded by wars (Table A3, columns [2]-[4].)22

Mechanisms. Two mechanisms could explain the observed incentive effect. If punishmentis more salient for those connected to the court-martialed officer, a guilty ruling may in-centivize related officers to exert more effort. Alternatively, a guilty ruling may weaken –

19These results are consistent with models of career concerns where connections and performance aresubstitutes (Jia et al., 2015; Xu, 2018).

20Many ties were formed even earlier. Since captains were responsible for manning their own ships, theyoften took the sons of friends and relatives on board at an early age. We have no systematic data on suchconnections.

21To ensure the magnitudes are comparable, we use the annual specification (Table A4, column [3]) tocompare the coefficients.

22Again, is unlikely that kins of guilty officers would be favored after the court-martial verdict.

13

or in the case of Admiral Byng even destroy entirely – the influence of an existing connec-tion. The loss of a powerful patron may thus induce officers to exert more effort in orderto distinguish themselves. To investigate the channels, we split the sample by whether thecourt-martialed officer is senior or junior to the officer he is connected with, as measuredby the years of seniority. If the increased performance is driven by the loss of a patron,we expect the observed effects to be driven by those officers whose guilty kin is indeedin a position of power. Consistent with the favoritism channel, the effect of a guilty court-martialed kin is concentrated among officers whose senior kin was convicted (column [4]).In contrast, officers with convicted junior kin do not perform significantly better (column[5]).

If the greater performance is driven by an attempt to compensate for a weakened socialtie, this incentive should be less for those who can draw on other, powerful connections.In columns [6]-[7], we split the sample by whether the officer is connected to the Admi-ralty in a given year – the apex of the naval hierarchy.23 As predicted, the incentive effectsare driven by officers not connected to the top of the Admiralty. In contrast, exposure toa court-martialed and guilty kin has no effect on performance of related officers who arethemselves directly connected to the Admiralty. Figure 2 shows the effect visually: whilethere is an increase in performance for officers with no ties to the Admiralty who see a kinconvicted, there is no effect for officers in the same situation, but with ties to the Admi-ralty.24

4 Conclusions

The Royal Navy in its heyday was a large and complex organization. While nominally runfrom the Admiralty in London, its ships sailed the seven seas, and often operated thou-sands of miles from home waters. Dispatches could take months to arrive. Under suchcircumstances, naval leaders in England were unable to control and direct naval warfaredirectly; instead, they had to rely on the judgement and motivation of men on the spot. Ac-cordingly, the Royal Navy used both rewards and punishments to align incentives (Allen,

23Following Voth and Xu (2020), the measure is a dummy that is 1 if the officer is either connected to theAdmiral of the Fleet or the First Lord Admiralty – the two highest ranked commanders – by less than 16degrees of separation (by consanguinity).

24We are unable to conduct the same sample cuts as before using Admiral Byng alone. Due to his excep-tional seniority, almost all connected officers are junior to him. As the fifth son of George Byng, 1st ViscountTorrington, Byng was also exceptionally well connected, thus providing too little variation.

14

2002).

Our evidence suggests that in settings with such severe principal-agent problems, strictpunishment can ”encourage others”. Performance in general did not improve after courtmartials – only convictions are associated with greater performance, and only those cap-tains related to the officer in question changed their behavior. This does not mean thatthere was no general effect of the navy’s strict rules, but they were not heightened by mak-ing punishment more salient in general.

There is a second, less obvious channel through which exemplary punishments can ”work”.By removing any chance of support through connections and patronage by a senior offi-cer, connected juniors were forced to prove their worth in battle – the positive effect onperformance is strongest for those who do not have other influential acquaintances, whocan substitute for the support of the court-martialed officer. Similar effects should be un-surprising in other organizations where informational asymmetries are large, and careersdepend on the support and mentoring of senior colleagues – such as in academia, banking,and the law.

ReferencesAger, P., L. Bursztyn, and H.-J. Voth (2016): “Killer Incentives: Status Competition and

Pilot Performance during World War II,” Tech. rep., National Bureau of Economic Re-search.

Allen, D. W. (2002): “The British Navy Rules: Monitoring and Incompatible Incentives inthe Age of Fighting Sail,” Explorations in Economic History, 39, 204–231.

——— (2011): The Institutional Revolution: Measurement and the Economic Emergence of theModern World, University of Chicago Press.

Baugh, D. A. (2008): “Byng, John (bap. 1704, d. 1757), naval officer,” in Oxford Dictionaryof National Biography, Oxford University Press.

Becker, G. S. (1968): “Crime and Punishment: An Economic Approach,” in The EconomicDimensions of Crime, Springer, 13–68.

Bentham, J. (1830): The Rationale of Punishment, R. Heward.Bhuller, M., G. B. Dahl, K. V. Løken, and M. Mogstad (forthcoming): “Incarceration

Spillovers in Criminal and Family Networks,” Journal of Political Economy.Blackstone, W. (1767/2017): Commentaries on the Laws of England: Book IV: On PublicWrongs, Jazzybee Verlag.

Chalfin, A. and J. McCrary (2017): “Criminal Deterrence: A Review of the Literature,”Journal of Economic Literature, 55, 5–48.

Chen, D. L. (2016): “The Deterrent Effect of the Death Penalty? Evidence from BritishCommutations during World War I,” mimeo.

Chen, M. K. and J. M. Shapiro (2007): “Do Harsher Prison Conditions Reduce Recidivism?A Discontinuity-based Approach,” American Law and Economics Review, 9, 1–29.

15

Donohue, J. and J. J. Wolfers (2006): “The Death Penalty: No Evidence for Deterrence,”The Economists’ Voice, 3.

Dull, J. R. (2007): The French Navy and the Seven Years’ War, U of Nebraska Press.Fehr, E. and S. Gachter (2002): “Altruistic Punishment in Humans,” Nature, 415, 137–140.Fehr, E. and B. Rockenbach (2003): “Detrimental Effects of Sanctions on Human Altru-

ism,” Nature, 422, 137–140.Gneezy, U. and A. Rustichini (2004): “Incentives, Punishment and Behavior,” Advances inBehavioral Economics, 572–89.

Hannay, D. (1914): Naval Courts Martial, University Press.Jia, R., M. Kudamatsu, and D. Seim (2015): “Political Selection in China: The Complemen-

tary Roles of Connections and Performance,” Journal of the European Economic Association,13, 631–668.

Katz, L., S. D. Levitt, and E. Shustorovich (2003): “Prison Conditions, Capital Punish-ment, and Deterrence,” American Law and Economics Review, 5, 318–343.

Kennedy, P. (2010): The Rise and Fall of the Great Powers, Vintage.Mocan, N. H. and R. K. Gittings (2006): “The Impact of Incentives on Human Behavior:

Can We Make It Disappear? The Case of the Death Penalty,” mimeo.Pope, D. (2013): Life in Nelson’s Navy, House of Stratus.Robson, M. (2015): A History of the Royal Navy: The Seven Years War, Bloomsbury Publish-

ing.Rodger, N. (1986): The Wooden World: An Anatomy of the Georgian Navy, Collins.——— (2005): The Command of the Ocean: A Naval History of Britain, 1649-1815, vol. 2, WW

Norton & Company.Voltaire, F. (1759/2005): Candide: Or Optimism, Penguin.Voth, H.-J. and G. Xu (2020): “Patronage for Productivity: Selection and Performance in

the Age of Sail,” mimeo.Xu, G. (2018): “The Costs of Patronage: Evidence from the British Empire,” American Eco-nomic Review, 108, 3170–98.

16

Figures

Figure 1: The execution of Admiral Byng and the performance of his kins

-.10

.1.2

.3.4

Perf.

diff

. Byn

g co

nnec

ted

vs. u

ncon

nect

ed

Jan 1

756

Jan 1

757

Jan 1

758

Jan 1

759

Notes: Reporting the performance gap between kins vs. non-kins of Admiral Byng in a balanced windowaround March 1757, the date Admiral Byng was executed. The coefficients are estimated using an augmentedversion of Table 2, Column [3] which allows the difference-in-differences to vary by month around the dayof execution. The figure plots 90% confidence intervals. Standard errors are clustered at the officer-level.

17

Figure 2: Performance of kins of guilty court-martialed officers, by ties to the Admiralty

-.2-.1

0.1

.2Pe

rform

ance

gap

: Con

nect

ed x

Gui

lty

-3 -2 -1 0 1 2 3Year around court-martial

(a) Ties to Admiralty

-.2-.1

0.1

.2Pe

rform

ance

gap

: Con

nect

ed x

Gui

lty

-3 -2 -1 0 1 2 3Year around court-martial

(b) No ties to AdmiraltyNotes: Reporting the performance gap between those connected vs. unconnected to guilty vs. acquitted court-martialed officers in a balancedthree year window around the year of court-martial, broken down by those who held no ties to the Admiralty (Panel a) and those who held tiesto the Admiralty (Panel b). Connected is defined as either holding kinship ties or having served together on a ship in the past. The coefficientsare estimated using an augmented version of Table 3, Columns [6]-[7] which allows the triple differences to vary by each year around thecourt-martial trial. Reporting 90% confidence intervals. Standard errors clustered at the officer × court-martial level.

18

Table 1: Officer characteristics - Byng kin vs. non-kin[1] [2] [3] [4] [5] [6]

Pre-execution Post-executionUnconnected Connected Diff. Unconnected Connected Diff.(N = 166) (N = 40) [2]-[1] (N = 166) (N = 40) [5]-[4]

Years of tenure 16.596 15.925 -0.671 18.584 17.925 -0.659(1.471) (1.471)

Post-captain 0.499 0.682 0.183** 0.798 0.941 0.142**(0.074) (0.061)

Years commanding 6.048 8.575 2.526*** 2.493 2.625 0.131(0.926) (0.170)

Ship guns 30.619 36.117 5.498* 43.281 47.546 4.264(3.124) (4.164)

Capture rate 0.188 0.151 -0.037 0.311 0.795 0.483***(0.072) (0.122)

Actions seen 0.057 0.116 0.058* 0.103 0.082 -0.020(0.031) (0.048)

Ships sunk 0.001 0.000 -0.001 0.006 0.016 0.010(0.001) (0.009)

Battle performance 0.155 0.185 0.030 0.385 0.875 0.490***(0.048) (0.121)

Notes: Comparing individual characteristics of officers by connectedness to Admiral Byng, before (pre1757) and after the execution (1757-1758). The sample comprises all serving officers who are activearound the execution of Admiral Byng (1756-1758). The years of tenure is the total number of years servedin the Royal Navy. Post-captain is a dummy that is 1 if the officer is a post-captain. Years commandingis the total years an officer has commanded a ship. Capture rate is the average number of enemy shipcaptures per year. Actions seen is the average number of actions participated per year. Ships sunk is theaverage number of enemy ships sunk per year. Battle performance is the mean number of enemy shipcaptures, ships sunk and actions participated (average battle performance). Finally, peerage is a dummyfor whether the officer is in the peerage dataset (a measure of eliteness). Standard errors are clusteredat the officer-level. *** p < 0.01, ** p < 0.05, * p < 0.1.

19

Table 2: Execution of Admiral Byng and the performance of his kins[1] [2] [3] [4] [5]

Battle performanceMean of dep. var 0.0285 0.0285 0.0285 0.0287 0.0417Post execution 0.003

(0.005)Byng kin -0.004 -0.004

(0.007) (0.008)Byng kin × Post execution 0.047*** 0.047*** 0.047*** 0.056*** 0.058***

(0.014) (0.014) (0.014) (0.019) (0.019)Year ×Month FEs Y Y Y YOfficer FEs Y Y YBaseline controls × Post Y YSample All naval officers PeerageObservations 7,416 7,416 7,416 7,344 1,944

Notes: Unit of observation is the officer × year × month. Balanced panel includes all officers servingbetween 1756 and 1758. Post execution is a dummy that is 1 after March 1757, the date Admiral Byng wasexecuted and 0 otherwise. Byng kin is a dummy that is 1 if the officer is connected to Admiral Byng byless than 16 degrees of separation (by consanguinity) and 0 otherwise. Baseline controls comprise pre-execution characteristics as shown in Table 1. These are (all computed up to 1756, the year before Byng’sexecution): the years of tenure; a dummy for whether the officer is a post-captain; the total years an officerhas commanded a ship; the mean number of enemy ship captures, ships sunk and actions participated(average battle performance); a dummy for whether the officer is in the peerage dataset (a measure ofeliteness). In Column [5], the sample excludes officers who are never connected by limiting the sampleto only those officers who are listed in the Peerage dataset. The dependent variable Battle performance isthe sum of enemy ship captures, enemy ships sunk, and the number of actions an officer participated inin a given month. Standard errors are clustered at the officer-level. *** p < 0.01, ** p < 0.05, * p < 0.1.

20

Table 3: The performance and ties to court-martialed officers[1] [2] [3] [4] [5] [6] [7]

Battle performanceMean of dep. var 0.301 0.179 0.179 0.191 0.174 0.146 0.324Kin × Post court-martial -0.034***

(0.012)× Guilty 0.066**

(0.029)Job tie × Post court-martial -0.050***

(0.013)× Guilty 0.093**

(0.038)Connected × Post court-martial -0.036*** 0.003 -0.056*** -0.019* -0.020

(0.007) (0.013) (0.008) (0.010) (0.017)× Guilty 0.059*** 0.050* 0.039 0.062** 0.015

(0.018) (0.029) (0.024) (0.024) (0.049)Court-martial FEs× Year around court-martial FEs Y Y Y Y Y Y Y× Officer FEs Y Y Y Y Y Y YTenure FEs Y Y Y Y Y Y YSample Pooled sample Court-martialed is Ties to Admiralty

Senior Junior No YesObservations 157,927 3,251,468 3,251,468 960,117 2,264,608 2,631,894 610,624

Notes: Unit of observation is the officer × court-martial × year. Sample includes all officers serving in athree year window around the year of each court-martial (balanced). Post court-martial is a dummy thatis 1 in the year of the court-martial and thereafter. Kin is a dummy that is 1 if the officer is connected to thecourt-martialed officer by less than 16 degrees of separation (by consangunity) and 0 otherwise. Job tieis a dummy that is 1 if the officer and the court-martialed officers have served on the same ship in thepast. Connected is a dummy that is 1 if the officer is either connected by kinship tie (Kin) or through a jobconnection (Job tie). Guilty is a dummy that is 1 if the court-martialed officer was judged to be guilty, and0 if the officer was acquitted. The dependent variableBattle performance is the combined number of enemyship captures, enemy ship sunk and actions participated in a given year. For column [1] the sample isrestricted to only officers who are found in the Peerage dataset (and for whom we can therefore computethe kinship ties). In Columns [4]-[5], the sample is divided by officers who are of lower seniority thanthe court-martialed officer (court-martialed is senior) and those who outrank the court-martialed officer(court-martialed is junior). In Columns [6]-[7], the sample is divided by officers who are connected tothe Admiralty (either to the Admiral of the Fleet or the First Lord Admiralty) by less than 16 degreesof separation (by consanguinity), and those who are unconnected to the Admiralty. Standard errors areclustered at the officer × court-martial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

21

Online appendix – not for publication

Figure A1: Fighting events over time

010

020

030

040

0

1700 1750 1800 1850

Fighting events Captures

Notes: Showing the total number of fighting events (captures, ships destroyed and actions participated)over time. Fighting events cluster around naval wars, and variation in captures drives the overall pattern.

22

Figure A2: Execution of Admiral Byng and kin performance - by degree of separation

0.0

25.0

5.0

75.1

Byng

kin

x P

ost e

xecu

tion

12 14 16 18 20Cut-off for connectedness to Byng

Notes: Estimating Table 2, column [3] using different cut-offs for maximum consanguinity to define Byngconnectedness. Reporting 90% confidence intervals. Standard errors are clustered at the officer-level.

23

Table A1: Performance results – Transformations and negative binomial regression[1] [2] [3] [4] [5] [6]

Battle performanceAny performance Inv. hyperb. sine Neg. binomial

Mean of dep. var 0.024 0.104 0.023 0.125 0.058 0.440Byng kin × Post execution 0.032*** 0.035*** 0.907**

(0.010) (0.010) (0.355)Connected × Post court-martial -0.016*** -0.023*** 0.027*

(0.003) (0.004) (0.015)Connected × Post court-martial × Guilty 0.018*** 0.031*** 0.164***

(0.007) (0.010) (0.042)Year ×Month FEs Y Y Y Y Y YOfficer FEs Y Y Y Y Y YSample Byng All Byng All Byng AllObservations 7,416 3,251,468 7,416 3,251,468 3,636 1,322,884

Notes: Unit of observation is the officer × year × month. In Columns [1], [3], and [5] the balancedpanel includes all officers serving between 1756 and 1758. In Columns [2], [4], and [6] the balancedpanel includes all officers serving around a three year window of each court-martial using the combinedkinship and on-the-job tie. Post execution is a dummy that is 1 after March 1757, the date Admiral Byngwas executed and 0 otherwise. Byng kin is a dummy that is 1 if the officer is connected to Admiral Byng byless than 16 degrees of separation (by consanguinity) and 0 otherwise. Connected is a dummy that is 1 ifthe officer is either connected by kinship tie (Kin) or through a job connection (Job tie). In Columns [1]-[2]the dependent variable is a dummy that is 1 if the officer captured, sunk or participated in any action ina given month. In Columns [3]-[4] the dependent variable is the inverse hyperbolic sine transformationof the combined number of enemy ship captures, enemy ships sunk, and the number of actions takenpart. The dependent variable Battle performance is the sum of enemy ship captures, enemy ships sunk,and the number of actions an officer participated in in a given month. Standard errors are clustered atthe officer-level. *** p < 0.01, ** p < 0.05, * p < 0.1.

24

Table A2: Performance of kins of court-martialed officers – AssignmentPanel A: Admiral Byng - family ties [1] [2] [3]

Command Guns PerformanceMean of dep. var 0.094 3.654 0.041Byng kin × Post execution 0.046 0.086 0.069***

(0.032) (0.132) (0.082)Year ×Months FEs Y Y YOfficer FEs Y Y YInteracted control Command

× Byng kinObservations 1,944 174 1,944Panel B: All court-martials [1] [2] [3]

Command Guns PerformanceMean of dep. var 0.460 3.601 0.179Connected × Post court-martial -0.034*** -0.030** -0.050***

(0.009) (0.013) (0.013)Connected × Post court-martial × Guilty 0.031 0.035 0.089**

(0.033) (0.052) (0.037)Court-martial FEs× Year around CM FEs Y Y Y× Officer FEs Y Y YTenure FEs Y Y YInteracted control Command

× Conn. × GuiltyObservations 3,251,468 1,462,239 3,251,468

Notes: Unit of observation is the officer × year × month in Panel A and officer × court-martial × yearin Panel B. Sample includes all officers serving between 1756 and 1758 (Panel A) and all officers servingaround a three year window of each court-martial using the combined kinship and on-the-job tie (PanelB). Post execution is a dummy that is 1 from 1757 onwards, the year Admiral Byng was executed and 0otherwise. Byng kin is a dummy that is 1 if the officer is connected to Admiral Byng by less than 16 degreesof separation (by consanguinity) and 0 otherwise. Kin is a dummy that is 1 if the officer is connectedto the court-martialed officer by less than 16 degrees of separation (by consanguinity) and 0 otherwise.Job tie is a dummy that is 1 if the officer and the court-martialed officers have served on the same ship inthe past. Connected is a dummy that is 1 if the officer is either connected by kinship tie (Kin) or througha job connection (Job tie). In column [1], the dependent variable is a dummy that is 1 if the officer wasgiven command over a ship. In column [2], the dependent variable is the (log) total number of gunsof the assigned ship. In column [3], the dependent variable Battle performance is the combined numberof enemy ship captures, enemy ship sunk and actions participated in a given year. Standard errors areclustered at the officer-level in Panel A, and the officer × court-martial level in Panel B. *** p < 0.01, **p < 0.05, * p < 0.1.

25

Table A3: Performance of kins of court-martialed officers – War-interacted(1) (2) (3) (4)

Battle performanceMean of dep. var 0.0285 0.300 0.179 0.179Byng kin × Post execution 0.069***

(0.020)Kin × Post court-martial -0.033***

(0.011)Kin × Post court-martial × Guilty 0.056**

(0.028)Job tie × Post court-martial -0.049***

(0.013)Job tie × Post court-martial × Guilty 0.093**

(0.037)Connected × Post court-martial -0.036***

(0.007)Connected × Post court-martial × Guilty 0.048***

(0.018)Year ×Months FEs YOfficer FEs YCourt-martial FEs× Year around CM FEs Y Y Y× Officer FEs Y Y YTenure FEs Y Y YInteracted control Total perf. × War × Kin War × Job War × Conn.

Byng kin × Guilty tie × Guilty × GuiltyObservations 1,944 157,927 3,251,468 3,251,468

Notes: Unit of observation is the officer × year × month in column [1] and officer × year in columns[2]-[4]. Sample includes all officers serving between 1756 and 1758 (Column [1]) and all officers servingaround a three year window of each court-martial (Columns [2]-[4]). Post execution is a dummy that is 1from 1757 onwards, the year Admiral Byng was executed and 0 otherwise. Byng kin is a dummy that is 1if the officer is connected to Admiral Byng by less than 16 degrees of separation (by consanguinity) and0 otherwise. Kin is a dummy that is 1 if the officer is connected to the court-martialed officer by less than16 degrees of separation (by consanguinity) and 0 otherwise. Job tie is a dummy that is 1 if the officerand the court-martialed officers have served on the same ship in the past. Connected is a dummy that is1 if the officer is either connected by kinship tie (Kin) or through a job connection (Job tie). In column[1], we interact the Byng kin dummy with the monthly total number of fighting events (combined num-ber of captures, enemy ships sunk and actions participated). In columns [2]-[4], we (triple) interact adummy that is 1 in a year of war with Kin, Post court-martial and Guilty (and job tie, and the combinedconnectedness measure, respectively). The dependent variable Battle performance is the combined num-ber of enemy ship captures, enemy ship sunk and actions participated in a given year. Standard errorsare clustered at the officer-level in column [1] and at the officer× court-martial-level in columns [2]-[4].*** p < 0.01, ** p < 0.05, * p < 0.1.

26

Table A4: Execution of Admiral Byng and the performance of his kins - Annualized data(1) (2) (3) (4)

Battle performanceMean of dep. var 0.283 0.283 0.283 0.371Post execution 0.403***

(0.047)Byng kin 0.040 0.042

(0.028) (0.028)Byng kin × Post execution 0.292** 0.288** 0.336*** 0.436**

(0.132) (0.131) (0.126) (0.199)Year FEs Y Y YOfficer FEs Y YSample All naval officers PeerageObservations 1,391 1,391 1,391 420

Notes: Unit of observation is the officer × year. Sample includes all officers serving between 1756 and1758. Post execution is a dummy that is 1 from 1757 onwards, the year Admiral Byng was executed and0 otherwise. Byng kin is a dummy that is 1 if the officer is connected to Admiral Byng by less than 16degrees of separation (by consanguinity) and 0 otherwise. In Column [4], the sample excludes officerswho are never connected by limiting the sample to only officers who are found in the Peerage dataset. Thedependent variable Battle performance is the combined number of enemy ship captures, enemy ship sunkand actions participated in a given year. Standard errors are clustered at the officer-level. *** p < 0.01, **p < 0.05, * p < 0.1.

27

Table A5: Officer characteristics - All court-martials - Connected vs. non-connected

[1] [2] [3] [4] [5] [6] [7] [8]Guilty (58 trials) Non-guilty (905 trials)

Unconnected Connected Diff. Unconnected Connected Diff. DiD(N = 23, 496) (N = 1, 927) [2]-[1] (N = 427, 533) (N = 12, 166) [5]-[4] [4]-[7]

Years of tenure Pre 18.522 20.143 1.621*** 18.150 19.746 1.596*** 0.025(0.267) (0.103) (0.284)

Post 22.486 24.115 1.629*** 22.098 23.724 1.625*** 0.003(0.267) (0.102) (0.281)

Post-captain Pre 0.595 0.790 0.194*** 0.592 0.783 0.191*** 0.002(0.011) (0.004) (0.012)

Post 0.681 0.873 0.192*** 0.682 0.866 0.184*** 0.008(0.010) (0.004) (0.011)

Years commanding Pre 1.266 1.530 0.264*** 1.385 1.590 0.205*** 0.058*(0.030) (0.012) (0.033)

Post 1.722 1.913 0.191*** 1.822 1.957 0.134*** 0.056(0.039) (0.015) (0.042)

Ship guns Pre 47.680 54.690 7.009*** 48.356 53.944 5.588*** 1.421*(0.694) (0.270) (0.746)

Post 49.912 59.267 9.354*** 50.643 57.696 7.053*** 2.301***(0.683) (0.261) (0.727)

Capture rate Pre 0.096 0.189 0.093*** 0.126 0.226 0.100*** -0.006(0.07) (0.003) (0.009)

Post 0.095 0.218 0.123*** 0.115 0.203 0.087*** 0.036***(0.006) (0.003) (0.008)

Actions seen Pre 0.036 0.066 0.029*** 0.053 0.093 0.040*** -0.010**(0.003) (0.001) (0.004)

Post 0.044 0.084 0.039*** 0.052 0.086 0.033*** 0.005(0.003) (0.001) (0.004)

Ships sunk Pre 0.003 0.006 0.002 0.003 0.006 0.003 -0.001(0.001) (0.001) (0.001)

Post 0.003 0.006 0.003 0.003 0.006 0.002 0.000(0.001) (0.001) (0.001)

Battle performance Pre 0.135 0.261 0.126*** 0.183 0.326 0.143*** -0.017(0.009) (0.004) (0.011)

Post 0.143 0.309 0.165*** 0.171 0.295 0.124*** 0.041***(0.008) (0.003) (0.009)

Notes: Unit of observation is the officer× court-martial event. Comparing average individual character-istics of officers by connectedness to court-martialed officer (guilty vs. non-guilty), before (3 years) andafter (3 years) the execution . The sample comprises all serving officers who are active around each ofthe 963 court-martial trials. The years of tenure is the total number of years served in the Royal Navy. Post-captain is a dummy that is 1 if the officer is a post-captain. Years commanding is the total years an officerhas commanded a ship. Capture rate is the average number of enemy ship captures per year. Actions seenis the average number of actions participated per year. Ships sunk is the average number of enemy shipssunk per year. Battle performance is the mean number of enemy ship captures, ships sunk and actionsparticipated (average battle performance). Finally, peerage is a dummy for whether the officer is in thepeerage dataset (a measure of eliteness). Standard errors are clustered at the officer-level. *** p < 0.01,** p < 0.05, * p < 0.1.

28

Table A6: Performance of kins of court-martialed officers – excl. acquitted for ship loss[1] [2] [3] [4] [5] [6] [7]

Battle performanceMean of dep. var 0.272 0.167 0.167 0.181 0.157 0.138 0.290Kin × Post court-martial -0.045**

(0.018)× Guilty 0.077**

(0.032)Job tie × Post court-martial -0.055**

(0.027)× Guilty 0.099**

(0.044)Connected × Post court-martial -0.003 0.043** -0.035** -0.020 -0.038

(0.011) (0.018) (0.015) (0.019) (0.029)× Guilty 0.026 0.010 0.018 0.064** 0.031

(0.020) (0.032) (0.027) (0.029) (0.054)Court-martial FEs× Year around court-martial FEs Y Y Y Y Y Y Y× Officer FEs Y Y Y Y Y Y YTenure FEs Y Y Y Y Y Y YSample Pooled sample Court-martialed is Ties to Admiralty

Senior Junior No YesObservations 63,441 638,920 638,920 271,323 361,971 513,667 123,648

Notes: Unit of observation is the officer × court-martial × year. Sample includes all officers serving in athree year window around the year of each court-martial (balanced), but is now restricted to the subsetof core court-martial trials and excluding the court-martials that resulted in an acquittal following theloss of a ship. Post court-martial is a dummy that is 1 in the year of the court-martial and thereafter. Kinis a dummy that is 1 if the officer is connected to the court-martialed officer by less than 16 degrees ofseparation (by consangunity) and 0 otherwise. Job tie is a dummy that is 1 if the officer and the court-martialed officers have served on the same ship in the past. Connected is a dummy that is 1 if the officeris either connected by kinship tie (Kin) or through a job connection (Job tie). Guilty is a dummy that is 1if the court-martialed officer was judged to be guilty, and 0 if the officer was acquitted. The dependentvariable Battle performance is the combined number of enemy ship captures, enemy ship sunk and actionsparticipated in a given year. For column [1] the sample is restricted to only officers who are found inthe Peerage dataset (and for whom we can therefore compute the kinship ties). In Columns [4]-[5], thesample is divided by officers who are of lower seniority than the court-martialed officer (court-martialedis senior) and those who outrank the court-martialed officer (court-martialed is junior). In Columns [6]-[7], the sample is divided by officers who are connected to the Admiralty (either to the Admiral of theFleet or the First Lord Admiralty) by less than 16 degrees of separation (by consanguinity), and thosewho are unconnected to the Admiralty. Standard errors are clustered at the officer× court-martial level.*** p < 0.01, ** p < 0.05, * p < 0.1.

29

Appendix

Extract from the Articles of War, Royal Navy, 1661-1866

A.1: Articles of War, No. 10 Every flag officer, captain and commander in thefleet, who, upon signal or order of fight, or sight of any ship or ships which itmay be his duty to engage, or who, upon likelihood of engagement, shall notmake the necessary preparations for fight, and shall not in his own person, andaccording to his place, encourage the inferior officers and men to fight coura-geously, shall suffer death or such other punishment, as from the nature anddegree of the offense a court martial shall deem him to deserve; and if any per-son in the fleet shall treacherously or cowardly yield or cry for quarter, everyperson so offending, and being convicted thereof by the sentence of a courtmartial, shall suffer death.

30